We introduce Adaptive Procedural Task Generation (APT-Gen), an approach to progressively generate a sequence of tasks as curricula to facilitate reinforcement learning in hard-exploration problems. At the heart of our approach, a task generator learns to create tasks from a parameterized task space via a black-box procedural generation module. To enable curriculum learning in the absence of a direct indicator of learning progress, we propose to train the task generator by balancing the agent's performance in the generated tasks and the similarity to the target tasks. Through adversarial training, the task similarity is adaptively estimated by a task discriminator defined on the agent's experiences, allowing the generated tasks to approximate target tasks of unknown parameterization or outside of the predefined task space. Our experiments on grid world and robotic manipulation task domains show that APT-Gen achieves substantially better performance than various existing baselines by generating suitable tasks of rich variations.

Label smoothing regularization (LSR) has a great success in training deep neural networks by stochastic algorithms such as stochastic gradient descent and its variants. However, the theoretical understanding of its power from the view of optimization is still rare. This study opens the door to a deep understanding of LSR by initiating the analysis. In this paper, we analyze the convergence behaviors of stochastic gradient descent with label smoothing regularization for solving non-convex problems and show that an appropriate LSR can help to speed up the convergence by reducing the variance. More interestingly, we proposed a simple yet effective strategy, namely Two-Stage LAbel smoothing algorithm (TSLA), that uses LSR in the early training epochs and drops it off in the later training epochs. We observe from the improved convergence result of TSLA that it benefits from LSR in the first stage and essentially converges faster in the second stage. To the best of our knowledge, this is the first work for understanding the power of LSR via establishing convergence complexity of stochastic methods with LSR in non-convex optimization. We empirically demonstrate the effectiveness of the proposed method in comparison with baselines on training ResNet models over benchmark data sets.

Adaptive gradient methods are typically used for training over-parameterized models capable of exactly fitting the data; we thus study their convergence in this interpolation setting. Under an interpolation assumption, we prove that AMSGrad with a constant step-size and momentum can converge to the minimizer at the faster $O(1/T)$ rate for smooth, convex functions. Furthermore, in this setting, we show that AdaGrad can achieve an $O(1)$ regret in the online convex optimization framework. When interpolation is only approximately satisfied, we show that constant step-size AMSGrad converges to a neighbourhood of the solution. On the other hand, we prove that AdaGrad is robust to the violation of interpolation and converges to the minimizer at the optimal rate. However, we demonstrate that even for simple, convex problems satisfying interpolation, the empirical performance of these methods heavily depends on the step-size and requires tuning. We alleviate this problem by using stochastic line-search (SLS) and Polyak's step-sizes (SPS) to help these methods adapt to the function's local smoothness. By using these techniques, we prove that AdaGrad and AMSGrad do not require knowledge of problem-dependent constants and retain the convergence guarantees of their constant step-size counterparts. Experimentally, we show that these techniques help improve the convergence and generalization performance across tasks, from binary classification with kernel mappings to classification with deep neural networks.

Readers are recommended to start with Zawacki-Richter et al.'s'Systematic review of research on artificial intelligence applications in higher education.' The authors reduced an initial trawl of 2656 articles published between 2007 and 2018 in peer reviewed journals down to 146 articles that met their selection criteria. The Zawacki-Richter at al. paper gives readers a good overview of the various areas where AI is being applied in higher education, as well as an indication of which areas researchers have tended to focus on. One of the areas identified by Zawacki-Richter et al. was the use of AI to predict final academic performance based on test results earlier in a course (profiling and prediction). The second paper in this issue by Akçapinar, Altun and Askar observed that 74% of the students who were unsuccessful at the end of term in an online computer science course in Turkey could be accurately predicted through the use of a specific algorithm (kNN) in as short as 3 weeks from the beginning of the course.

Virtual reality and artificial intelligence are being touted as the transformative tools that will reduce human error in the operating room. But how do they work? We speak to Gabriel Jones, CEO and co-founder of Seattle-based Proprio, to find out. Proprio was founded in 2016 by Dr. Sam Browd, a pediatric neurosurgeon, with the initial goal to eliminate the need for loupes - the magnifying glasses surgeons wear to perform delicate operations, and replace them with a digital alternative. In the four years since, Browd and his co-founders Gabriel Jones and James Youngquist (Chief Technology Officer) have added machine learning, computer vision, robotics and mixed reality to their solution to augment human vision during surgery.

Every day there is something new going on in the world of AWS Machine Learning--from launches to new use cases to interactive trainings. Check back at the end of each month for the latest roundup. This month we announced native support for TorchServe in Amazon SageMaker, launched a new NFL Next Gen Stat, and enhanced our language services including Amazon Transcribe and Amazon Comprehend. TorchServe is now natively supported in Amazon SageMaker as the default model server for PyTorch inference to help you bring models to production quickly without having to write custom code. Want more news about developments in ML? Check out the following stories: Laura Jones is a product marketing lead for AWS AI/ML where she focuses on sharing the stories of AWS's customers and educating organizations on the impact of machine learning.

I started Practical Deep Learning for Coders 10 days ago. I am compelled to say their pragmatic approach is exactly what I needed. I started data science by learning Python, Pandas, NumPy, and whatever I needed in a short few months. I did whatever courses I need to do (e.g. Kaggle micro-courses) and whatever books I needed to read (e.g.

"Neural Networks From Scratch" is a book intended to teach you how to build neural networks on your own, without any libraries, so you can better understand deep learning and how all of the elements work. This is so you can go out and do new/novel things with deep learning as well as to become more successful with even more basic models. This book is to accompany the usual free tutorial videos and sample code from youtube.com/sentdex. This topic is one that warrants multiple mediums and sittings. Having something like a hard copy that you can make notes in, or access without your computer/offline is extremely helpful.

Roger Perlmutter, the head of research and development at Merck, is joining the board of Insitro, a firm focused on using artificial intelligence to discover drugs. Insitro, backed with $243 million in venture capital from firms including Casdin Capital and ARCH Venture Partners, was founded by Daphne Koller, known for co-founding Coursera, the online learning firm, and working at Calico, a drug discovery arm of Alphabet. The company has a research partnership with Gilead Sciences. Unlock this article by subscribing to STAT Plus and enjoy your first 30 days free! STAT Plus is STAT's premium subscription service for in-depth biotech, pharma, policy, and life science coverage and analysis.

Approximately 45 US states do not require schools to teach students handwriting, but a new study suggests the skill is vital to a child's development. Following an examination of brain activity, researchers found using a pen and paper helps children learn more and remember better than if they record information on a computer. The data showed an increase of activity in the sensorimotor parts of the brain, which is involved with processing, attention and language. Scientist also found that the act is beneficial for adults, suggesting they will remember contents better after writing them down. The research was conducted by a team at Norwegian University of Science and Technology (NTNU), who now suggest national guidelines need to ensure children are receiving some handwriting lessons.